regarding giving number of nodes
at present i'm meshing an airfoil using icemcfd,i'm confused how many number of nodes should be given to a edge.i know that increase the number of nodes increases computational time and increases accuracy .but when we run it in fluent for analysis we do not know how many iterations we have to use..does the number of nodes affect the number of iterations, if so how??
and how we can optimize the number of nodes so that the number of iterations to be done are low??
The solver adjusts the solution and re-solves until the solution doesn't change very much between iterations... These changes are often referred to as residuals, and when they are small enough, we say the model has converged.
Having a very fine mesh will allow for more accuracy, but that will also allow you to capture interesting turbulence, swirl, etc. that may actually slow down convergence because it takes the solver longer to work it all out or because the turbulence is unstable (no steady/stable solution).
Having too coarse of a mesh can reduce your accuracy and, in many cases, also make it difficult for your model to converge ;^)
But you shouldn't be trying to size your mesh for convergence or a number of iterations... The goal should be accuracy within tolerances. To work that out, most mature simulation groups have done refinement studies in order to learn what level of refinement is needed (low end) and what is wasteful (high end). This actually requires meshing a model at a variety of resolutions and comparing those solutions to a test case.
As for iterations, you should set your convergence criterion and solve. If your model reaches the full number of iterations set, you need to run it for more iterations. The stop should be based on the convergence of the residuals, not the number of iterations.
i got it now ..
thanks for the reply simon..
|All times are GMT -4. The time now is 00:47.|